Executable flexibility coordination of electric vehicle aggregators via temporal reinforcement learning in coupled electricity markets

Electric vehicle aggregators are expected to provide flexibility across multiple electricity market layers. However, advance commitments may become infeasible when mobility and service activation change the state-of-charge trajectory. This paper proposes an executable decision framework for electric vehicle aggregators in coupled electricity markets. The framework first converts nominal electric vehicle flexibility into executable capacity. It links traffic-induced state-of-charge evolution with charging and discharging power limits. A coordinated bidding mechanism is then developed. The mechanism allocates executable capacity between day-ahead and real-time markets. Price errors, reserve activation, electric vehicle availability, and deviation penalties are embedded in bidding decisions. A temporal learning strategy is further designed to capture inter-stage dependencies under evolving feasible decision ranges. The proposed method is validated through benchmark simulations and a large-scale real-world case study. Results demonstrate that the method improves executability consistency by 6.2% and reduces distribution-system operating costs by 4.8%. The revenue of electric vehicle aggregators increases by 14.75% over energy-only participation and 8.55% over energy-reserve participation. In a Tianjin real-world case, the average revenue gain reaches 15%-20%. This study supports the transition of electric vehicle aggregation from flexibility potential estimation to executable market coordination.